MAY 2020 A R C H E R E T A L . 1509

Observations of Submesoscale Variability and Frontal Subduction within the Mesoscale Eddy Field of the Tasman Sea

MATTHEW ARCHER,AMANDINE SCHAEFFER,SHANE KEATING, AND MONINYA ROUGHAN School of Mathematics and Statistics, University of New South Wales, Sydney, New South Wales, Australia

RYAN HOLMES School of Mathematics and Statistics, and Climate Change Research Centre, and Centre of Excellence for Climate Extremes, University of New South Wales, Sydney, New South Wales, Australia

LIA SIEGELMAN California Institute of Technology, Pasadena, California

(Manuscript received 28 May 2019, in final form 22 March 2020)

ABSTRACT

Submesoscale lenses of water with anomalous hydrographic properties have previously been observed in the East Australian Current (EAC) system, embedded within the thermocline of mesoscale anticyclonic eddies. The waters within these lenses have high oxygen content and temperature–salinity properties that signify a surface origin. However, it is not known how these lenses form. This study presents field observations that provide insight into a possible generation mechanism via subduction at upper-ocean fronts. High- resolution hydrographic and velocity measurements of submesoscale activity were taken across a front between a mesoscale eddy dipole downstream of the EAC separation point. The front had O(1) Rossby number, strong vertical shear, and flow conducive to symmetric instability. Frontogenesis was measured in conjunction with subduction of an anticyclonic water parcel, indicative of intrathermocline eddy formation. Twenty-five years of satellite imagery reveals the existence of strong mesoscale strain coupled with strong temperature fronts in this region and indicates the conditions that led to frontal subduction observed here are a persistent feature. These processes impact the vertical export of tracers from the surface and dissipation of mesoscale kinetic energy, implicating their importance for understanding regional ocean circulation and biological productivity.

1. Introduction Munk et al. 2000), it was not until recently that the global extent, complexity, and importance of this scale Over the past decade, a succession of numerical and class has been realized (Mahadevan 2016; Su et al. 2018; observational studies has revealed the rich tapestry of Torres et al. 2018). Submesoscales provide a pathway for submesoscale ocean currents that have horizontal scales the cascade of energy from the large scale, where it is of O(0.1–10) km and evolve and decay much faster than input by atmospheric forcing, to the microscale, where it their mesoscale counterparts [e.g., Capet et al. 2008; is dissipated, and in modulating the mesoscale field Thompson et al. 2016; see McWilliams (2016) for a re- (Molemaker et al. 2010). They are also associated with view]. While submesoscale ocean features were ob- enhanced turbulence, lateral/vertical mixing, and strong served several decades earlier (e.g., Flament et al. 1985; vertical velocities (Klymak et al. 2016; D’Asaro et al. 2011; Shcherbina et al. 2015) and therefore influence primary Supplemental information related to this paper is available at productivity and atmospheric exchange (Mahadevan 2016; the Journals Online website: https://doi.org/10.1175/JPO-D-19- Omand et al. 2015). 0131.s1. Submesoscale currents form through a variety of mech- anisms, including mixed layer instabilities (Boccaletti et al. Corresponding author: Matthew Archer, matthew.robert.archer@ 2007), mesoscale frontogenesis (Spall 1997), and topo- gmail.com graphic wakes (Molemaker et al. 2015). As such, they

DOI: 10.1175/JPO-D-19-0131.1 Ó 2020 American Meteorological Society. For information regarding reuse of this content and general copyright information, consult the AMS Copyright Policy (www.ametsoc.org/PUBSReuseLicenses). Unauthenticated | Downloaded 10/06/21 12:07 PM UTC 1510 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 50 exhibit strong spatial and temporal inhomogeneity with In addition to frontal instability, other mechanisms for considerable variation both geographically and season- ITE generation include bottom drag on boundary cur- ally (Sasaki et al. 2014; Callies et al. 2015; Rocha et al. rents at topographic features (e.g., D’Asaro 1988; Gula 2016). Difficulties in measuring these intermittent small- et al. 2019), diapycnal mixing and geostrophic adjust- scale features have restricted observational studies to a ment (e.g., McWilliams 1985), and baroclinic instabil- limited number of localized regions, via ship-based field ity of an undercurrent (e.g., Jungclaus 1999). Despite campaigns [e.g., CARTHE (Poje et al. 2014), OSMOSIS modeling studies that can explain the ITE generation (Thompson et al. 2016), SMILES (Adams et al. 2017), mechanism via frontal instability and subduction (Spall LatMix (Shcherbina et al. 2015), repeat ship-of-opportunity 1995; Thomas 2008), there are few direct observations of sampling (Wang et al. 2010; Qiu et al. 2017; Chereskin this process (Lee et al. 2006). However, such a mecha- et al. 2019), or coast-based high-frequency (HF) radar nism may play an important role in ocean–atmosphere (Kim et al. 2011; Archer et al. 2015; Schaeffer et al. 2017; exchange, by exporting near-surface water masses Yoo et al. 2018)]. Therefore, many regions of the world influenced by atmospheric forcing down into the ocean remain unsampled with respect to submesoscale ocean interior (Klein and Lapeyre 2009). variability. In the Tasman Sea east of Australia, Baird and Ridgway Many of these studies have highlighted the critical (2012, hereafter BR12) identified submesoscale lenses of role of frontal zones in transferring kinetic and poten- anomalous water at depth, embedded within mesoscale tial energy from the mesoscale current field to sub- anticyclonic eddies. The water mass characteristics of these mesoscales. Fronts can form by the mesoscale stirring of lenses revealed their origin as Bass Strait water, but BR12 lateral density gradients; frontogenesis occurs when the did not observe how they form. Brannigan et al. (2017), density gradients are sharpened through lateral strain. based on results of high-resolution idealized simulations, As the front sharpens, an ageostrophic secondary cir- conjectured that the Bass Strait lenses observed by BR12 culation can develop to drive the alongfront flow back to mayhaveformedduetosubmesoscale instabilities in the thermal wind balance through restratification (Hoskins mixed layer of mesoscale eddies, but no observations of 1982). This takes the form of a single overturning cell this process have been made. that can drive vertical exchange between the mixed Here we present high-resolution in situ observations of layer and interior. Submesoscale instabilities can feed submesoscale processes in the Tasman Sea (Fig. 1)taken off the local potential and kinetic energy of the inten- during a research cruise in austral late winter/early spring sifying front. These instabilities are strongly influenced 2017. The measurements were made across a frontal zone by the atmosphere—air–sea heat fluxes and frictional separating an anticyclonic warm core eddy and cy- wind forcing can both act to further destabilize the water clonic cold core eddy, immediately south of the East column. The particular flavor of submesoscale instabil- Australian Current (EAC) separation zone (Fig. 1). Our ity that dominates is dependent on the distributions analysis indicates the presence of frontogenesis-induced of vorticity and density gradients, which can incite submesoscale subduction, symmetric instability, and the gravitational instability (Haine and Marshall 1998), formation of an intrathermocline eddy. These observa- centrifugal/inertial instability (Molemaker et al. 2015), tions lend support to the theory that such submesoscale symmetric instability (Thomas et al. 2013), and mixed lenses may form regularly between mesoscale eddies in layer baroclinic instability (Stone 1970; Boccaletti et al. the Tasman Sea due to submesoscale instabilities in the 2007; Fox-Kemper et al. 2008). mixed layer. During frontal instability, surface water can be sub- ducted along isopycnals, develop anticyclonic vertical 2. Study region vorticity, and separate from the mixed layer to form an intrathermocline eddy (hereafter ITE) [Spall 1995; Ocean circulation in the Tasman Sea is dominated by Thomas 2008; also termed a submesoscale coherent the presence of the EAC, the western boundary current vortex (McWilliams 1985)]. ITEs are submesoscale (WBC) of the South Pacific subtropical gyre (Suthers anticyclonic eddies observed throughout the stratified et al. 2011; Oke et al. 2019). The EAC is known for interior of the ocean (Dugan et al. 1982; Kostianoy and being the most variable of WBCs globally, with a large Belkin 1989). They are characterized by their small eddy kinetic energy (EKE) to kinetic energy (KE) ratio length scales, well-mixed and anomalous water proper- (Godfrey et al. 1980; Archer et al. 2017). The EAC flows ties, long lifetimes, and large propagation distances. For along the continental slope of east Australia until it these reasons, the cumulative effect of ITEs in the separates from the coast typically between 308 and world’s oceans has a potentially significant impact on the 328S(Cetina-Heredia et al. 2014), flowing eastward as transport and distribution of heat, salt, and other tracers. the Tasman Front. At the separation point, mesoscale

Unauthenticated | Downloaded 10/06/21 12:07 PM UTC MAY 2020 A R C H E R E T A L . 1511

FIG. 1. Satellite imagery in the Tasman Sea reveals the rich submesoscale tapestry of filaments and eddies present around, and within, the mesoscale eddy field: (a) SST (8C) and (b) chlorophyll a concentration using the OC3 2 algorithm (mg m 3). Both datasets are processed to level L2, with a spatial resolution of 750 m, obtained on 4 Sep 2017 from an ascending swath at approximately 0330 UTC from Suomi-NPP VIIRS. Data source: http:// oceandata.sci.gsfc.nasa.gov. The inset figure in (a) shows the continent of Australia; the red box denotes the study area, and the yellow box denotes Bass Strait, which separates Tasmania from mainland Australia. The East Australian Current (EAC) is denoted by the large gray arrow, and the eddy dipole is denoted with the white (cyclonic eddy) and black (anticyclonic eddy) arrows. anticyclonic eddies intermittently shed from the EAC distributions of vorticity and divergence. They occur retroflection and interact with cyclonic eddies from intermittently all year round, and are associated with offshore (Fig. 1). Local wind forcing has been shown to upwelling and elevated chlorophyll a concentrations influence the timing of EAC eddy shedding events, (Roughan et al. 2017; Schaeffer et al. 2017). Archer et al. whereas remote wind forcing appears to have limited (2018) used over 4 years of high-resolution (1.5 km) HF impact on this intrinsically variable WBC system (Bull radar observations in this upstream region to explore et al. 2017). The area immediately south of where the surface current variability; in EKE wavenumber spectra 2 EAC separates has been termed ‘‘eddy avenue’’ for its they identified a k 3 power spectral slope, which indicates abundance of energetic mesoscale eddies with excep- the dominance of mesoscale motions, and little-to-no tionally large anomalies in sea surface temperature seasonality in submesoscale variance. Poleward of where (SST) and chlorophyll a (Everett et al. 2012). For this the EAC separates from the coast, satellite imagery re- reason, the focus of earlier studies in the region was to veals submesoscale features (Fig. 1), yet there have been characterize the variability and dynamics of mesoscale no dedicated in situ observations to determine the sub- eddies, from a statistical perspective as well as individual surface hydrographic and velocity structure of these case studies (Macdonald et al. 2013; Everett et al. 2015; scales. The measurements analyzed here are the first to Roughan et al. 2017; Oke et al. 2019). focus on these upper-ocean submesoscales poleward of Several recent studies have investigated submesoscale the EAC separation. variability of the EAC as it flows along the continental shelf upstream of its separation (Roughan et al. 2017; Schaeffer et al. 2017; Mantovanelli et al. 2017; Archer 3. Data and methods et al. 2018). Submesoscale flow in this regime is char- a. Shipboard observations acterized by instability of the meandering EAC and generation of frontal eddies (Schaeffer et al. 2017). The primary observations used in this study were These eddies have O(1) Rossby number, and asymmetric collected during a research cruise in the Tasman Sea,

Unauthenticated | Downloaded 10/06/21 12:07 PM UTC 1512 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 50 austral late winter/early spring (31 August–18 September by linearly interpolating between successive up- and 2017), aboard the R/V Investigator (Malan et al. 2020). downcasts. Standard shipboard underway measurements used in- b. Satellite imagery clude salinity and temperature from a thermosalino- graph (Sea-Bird SBE 21) with measurement depth of To characterize the mesoscale eddy field, we use 2mwithinthedropkeel,andwindspeedanddirection satellite-measured absolute dynamic topography (ADT) from a wind vane (RM Young Wind Sensor Type and geostrophic currents obtained from the DUACS- 05108) on the starboard foremast. Vertical profiles of DT2014 L4 gridded multimission altimeter daily prod- conductivity (SBE4C), temperature (SBE3T), pres- uct on a 0.25830.258 resolution grid (Pujol et al. sure, and dissolved oxygen (DO; SBE43) were made 2016). Backward-in-time finite-size Lyapunov expo- 2 by a Sea-Bird SBE 911, down to a depth of 1000 m at 1- nents (FSLE; day 1) are supplied by AVISO at a dbar intervals. The voyage data are quality controlled horizontal resolution of 0.048, calculated from geo- by the Australian Marine National Facility staff using strophic velocities. FSLEs are a Lagrangian diagnostic their standard procedures (https://doi.org/10.4225/08/ measuring the exponential rate of separation lbetween 5a964b9041903). Potential density, Absolute Salinity, two neighboring particles during a time of advection t, 21 Conservative Temperature, and the thermal expansion defined as l 5 t log(d0/df), where d0 (df)istheinitial coefficient were calculated using the TEOS-10 toolbox (final) separation distance (d’Ovidio et al. 2004).

(McDougall and Barker 2011). AVISO-defined parameters are d0 5 0.028 and df 5 Vertical profiles of horizontal currents were collected 0.68, with a maximum integration window of 200 days. down to a depth of ;800 m with 8-m vertical bins FSLEs are used here to diagnose the strain field and 1.2-km horizontal spacing, using a ship-mounted (Waugh and Abraham 2008) and are positively corre- acoustic Doppler current meter (S-ADCP; RDI OS lated with gradients of buoyancy at the submesoscale, as 75-kHz narrowband operation). The University of Hawaii’s shown in the Antarctic Circumpolar Current (ACC) Common Ocean Data Access System (CODAS) was used (Siegelman et al. 2020). Since we use several satellite for postprocessing, including corrections for ship speed SST and chlorophyll a datasets, we note the data source andheading,withdataoutputas5minaveragesfrom when used in the text. 1-Hz sampling frequency. Unless otherwise stated, all c. Diagnostic calculations velocity is rotated to along the ship track (u component) and across track (y component), which is approximately The satellite data are on a Cartesian longitude– equivalent to being across front (u) and alongfront (y), latitude grid, so we calculate both horizontal gradient since the ship passed the frontal zone almost perpendic- terms (›/›x and ›/›y). However, for in situ measure- ular to its axis (within 48). We tested the sensitivity of our ments, horizontal gradients can only be calculated results to this rotation angle, comparing velocity data along the ship track. Therefore, we orientate the co- rotated to the surface front velocity orientation as well as ordinate system along track (x direction) and across the 200-m depth-averaged front orientation; the results track (y direction), and assume that across-front gra- are not sensitive to these changes. dients (›/›x) are significantly larger than alongfront A Triaxus (a vertically profiling towed body) was gradients (›/›y / 0), which is a reasonable approxi- deployed to map the hydrographic structure across the mation based on the high-resolution SST imagery front at high resolution (at 1446–2227 UTC 12 September). depicting a nearly straight front during the in situ It was equipped with a dual-sensor Seabird SBE 911, sampling (see section 5). The z coordinate is positive collecting temperature, conductivity, pressure, and DO. upward. We calculate all gradients using a centered Chlorophyll was measured by an Eco-Triplet payload. second-order differencing scheme. 2 The Triaxus was towed at a speed of 8 kt (;4ms 1), 1) IN SITU VELOCITY GRADIENT TENSOR approximately 1.5 km behind the vessel, in a sawtooth pattern with horizontal spacing of ;2km between From the horizontal velocity gradient tensor, we cal- bottom turns, over vertical dives of ;200 m. This culate vertical relative vorticity equates to a time lag between the S-ADCP and Triaxus ›y measurements of approximately 6 min. The depth of z 5 , (1) the dives was progressively increased throughout the ›x deployment to capture the base of the mixed layer, as horizontal divergence the ship moved from the cold core eddy to the warm core eddy. All Triaxus-measured variables are gridded ›u d 5 , (2) horizontally (dx ; 1 km) and vertically (dz 5 1m) ›x

Unauthenticated | Downloaded 10/06/21 12:07 PM UTC MAY 2020 A R C H E R E T A L . 1513 and strain ›y y ›b 1 ›b 2 q 5 f 1 u 1 u 2 , (9) " # cyl ›r R ›z f ›r 1/2 ›u 2 ›y 2 s 5 1 . (3) where y is the radial velocity, r is the radial coordinate, ›x ›x u and R is the radius of curvature.

3) ATMOSPHERIC FORCING 2) IN SITU POTENTIAL VORTICITY CALCULATION Frictional forces associated with wind forcing over Ertel potential vorticity (hereafter PV, or denoted by an ocean front can modulate the PV (and so poten- the symbol q) can be used as a diagnostic to indicate the tially drive instability) as captured by the Ekman susceptibility of the flow to instability (e.g., Hoskins buoyancy flux: 1974; Thomas et al. 2008). In its full form it is t y ›b EBF 5 , (10) q 5 (f 1 = 3 u) =b, (4) r › 0 f x z50 5 u y w f where u ( , , ) is the velocity vector, is the Coriolis where ty is the alongfront component of the wind 52 0 parameter, and b gr /r0 is the buoyancy (g is ac- stress. Winds directed along the front in the direction 0 celeration due to gravity, r is a density perturbation, of the thermal wind shear (i.e., downfront, EBF . 0) and r0 is a reference density, taken here as the full-depth will drive a frictional Ekman transport that forces mean). When PV has the opposite sign of f (fq , 0), the dense water over light, acting to steepen the front flow is inherently unstable (Thomas et al. 2013). PV can and reduce the vertical stratification (Thomas and be decomposed into two main terms: Lee 2005), injecting positive PV (in the Southern Hemisphere) into the ocean, encouraging fq , 0. 5 1 q qvert qbc , (5) The interaction of wind stress with the relative vor- ticity of the ocean front can also induce Ekman 5 1 2 qvert (f z)N , and (6) convergence/divergence and drive vertical motions (Thomas and Lee 2005). Thus, independent of any ›y ›b ›y 5 2 52 2 air–sea heat and salt fluxes, downfront winds are qbc M , (7) ›z ›x ›z destabilizing. For comparison with previous studies, we convert EBF into an equivalent heat flux (e.g., where N2 5 ›b/›z is the vertical buoyancy gradient, and Thompson et al. 2016): M2 5 ›b/›x is the horizontal buoyancy gradient. The term qvert relates to the vertical component of absolute r0Cp 1 2 Q 52 EBF, (11) vorticity (f z) and N , and qbc relates to the horizontal EK ag relative vorticity (vertical shear) and M2. Gravitational 2 , instability can occur when N 0. If the water column is where Cp is the specific heat of seawater (taken here 2 21 21 stably stratified (N . 0) then for fqvert , 0, inertial or as 3992 J kg 8C ), and a is the thermal expansion centrifugal instability may occur. However, if fq , 0is coefficient, a function of salinity and temperature. EBF due to the baroclinic term, fqbc , 0, then symmetric used here is calculated using the 2-m density gradient instability can occur (Thomas et al. 2013). When the derived from the underway measurements. A negative frontal flow is approximately in thermal wind balance QEK (associated with downfront winds, EBF . 0) de- (i.e., the vertical shear is geostrophic ›yg/›z 5 1/f(›b/›x), stabilizes the mixed layer. qbc can be expressed as (Thomas et al. 2008) The total air–sea heat flux is QHF 5 Qshortwave 1 1 1 1 22 Qlongwave Qlatent Qsensible Qrain (W m ). The term 2 2 4 ›y 1 ›b M QHF is calculated from shipboard meteorological and qg 52f g 52 52 . (8) bc ›z f ›x f oceanographic observations using the COARE 3.0a bulk algorithms (Fairall et al. 1996). When QHF is pos- g Thus fqbc is a negative definite quantity so that the baro- itive (negative), the ocean gains (loses) heat, stabilizing clinicity of the thermal wind shear always acts toward (destabilizing) the water column. fq , 0 (i.e., the horizontal vorticity of the geostrophic 4) FRONTOGENESIS shear is ‘‘anticyclonic’’). Therefore, symmetric instabil- g ity can occur when the magnitude of qbc exceeds qvert. The frontogenesis function provides a measure of the To incorporate the effect of flow curvature, we calculate rate at which horizontal velocity gradients can increase PV in cylindrical coordinates (Shakespeare 2016): the lateral buoyancy gradient (Hoskins 1982),

Unauthenticated | Downloaded 10/06/21 12:07 PM UTC 1514 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 50 ›u ›b ›b an offshore cyclonic eddy that propagated westward. F 5 2 2 , (12) ›x ›x ›x As the anticyclonic eddy detached from the EAC jet (1–10 August), the cyclonic eddy squeezed in between where F is one term in the equation for the Lagrangian the two warm water masses, thus forming an eddy evolution of the horizontal buoyancy gradient (Holmes dipole with the anticyclone, southwest of the EAC et al. 2014)(D/Dt)j›b/›xj2 and is calculated using the jet separation [see Malan et al. (2020) for details on Triaxus-measured buoyancy gradient and ADCP-measured the eddy-driven cross-shelf transport of this dipole]. velocity (Thomas et al. 2013). Parameter F . 0 indi- Over this period the anticyclonic eddy slowly decayed, cates that horizontal strain is acting to drive fronto- with decreasing ADT amplitude and spatial scale. genesis. Typically, this will induce an ageostrophic The cyclonic eddy also shrunk as it moved westward secondary circulation that partially balances the fron- and became constrained between the EAC and anti- togenetic tendency (through differential vertical ad- cyclonic eddy. Toward the end of September, the vection; Hoskins 1982). cyclonic eddy was squeezed out to the south as the d. Data processing anticyclonic eddy reconnected with the EAC jet (Figs. 2 and 3). 1) COLLOCATING IN SITU DATASETS b. Mesoscale frontal structure To calculate specific terms (see section 3c), the hydro- The evolution of the eddy dipole generated strong graphic and velocity observations need to be collocated. fronts that divided the distinct water masses of the ADCP-measured velocity and its vertical/horizontal shear warm EAC, cold cyclonic eddy, and warm anticy- is linearly interpolated from its native grid to the clonic eddy (Fig. 3a). The fronts exhibit large gradi- Triaxus grid in x/z coordinates. The final dataset has a ents in SST over short distances (with a negative resolution of 1 m in the vertical and ;1 km along track. gradient within the eddy dipole ;348–358S; Fig. 3b). We also tested interpolating from the Triaxus to ADCP The time–latitude Hovmöller plot of FSLE (Fig. 3c) grid, with the same results. Horizontal and vertical shows the evolution over a 2-month period. FSLEs 2 gradients in all properties are computed via central reach values of up to 1 day 1, characteristic of an in- differencing. tense strain field and similar to values found in the ACC (Siegelman et al. 2020). As the cyclonic eddy 2) CALCULATING ERROR BARS moved westward between the EAC and the anticy- To evaluate the significance of the PV field, given the clonic eddy, strain increased, and FSLE values be- limited observations available, we run a Monte Carlo tween the two eddies were intensifying as the ship 2 experiment using error estimates of velocity (5 cm s 1), passes the front. This indicates the front is undergoing temperature (0.0028C) and salinity (0.005 psu). These frontogenesis. estimates are twice the manufacturer-supplied accuracy High-resolution SST imagery of the front on the day for temperature and salinity, and twice the error found of sampling shows it to have slight anticyclonic curva- in an empirical study of the RDI S-ADCP accuracy ture (Figs. 4a,b). To calculate the radius of curvature R, (Fischer et al. 2003). We assume Gaussian error distri- we best fit a circle to this frontal region, obtaining R 5 butions with zero mean and STD equal to the estimated 80 km. The cyclogeostrophic Rossby number, C 5 yu/fR, error, then randomly assign these errors to each grid provides a metric of the influence of curvature on the point. From these new noisy fields, we calculate gradi- flow (Shakespeare 2016). Based on a depth-averaged ents and the terms described in section 3c; then repeat yu at the front, and the R calculated above, we obtain 1000 times. From these one thousand realizations, we C 520.14, which is relatively low and indicates a derive confidence intervals based on the Student’s t weak influence. The negative value is due to anticy- distribution, assuming a sample STD (unknown pop- clonic curvature and implies an increase in yu for the ulation STD), with N 2 1 degrees of freedom. forces to remain in balance (horizontal pressure gra- dient, Coriolis, and centripetal). 4. Mesoscale eddy field (austral winter 2017) At the time of in situ sampling, the ADT-derived geostrophic speed of the onshore jet of the dipole a. EAC eddy shedding event 2 reached a maximum of 1.3 m s 1 (Fig. 4f). The geo- During August and September 2017, satellite ADT strophic vorticity (normalized by jfj) measured nearest maps reveal an EAC anticyclonic eddy shedding and the front reached 20.15f, while the anticyclonic vorticity reattachment event at approximately 348S(Fig. 2). was 0.08f. Within the eddy cores, vorticity values peaked The anticyclonic eddy shedding was associated with at 20.4f and 0.3f, respectively, indicating relatively high

Unauthenticated | Downloaded 10/06/21 12:07 PM UTC MAY 2020 A R C H E R E T A L . 1515

FIG. 2. Evolution of the EAC eddy shedding event shown in nine snapshots of absolute dynamic topography (ADT), starting 1 Aug when an anticyclonic eddy was shed from the EAC through to 30 Sep, when the same feature reattached to the EAC. On 12 Sep, in situ measurements were taken along the ship transect denoted by the black line. The eddy dipole feature is encased within the blue box. The black dashed line in the top-right figure denotes the transect over which the Hovmöllers in Fig. 3 are plotted. mesoscale Rossby numbers. Even though these geo- the mesoscale field (e.g., Fig. 4g). At the shallowest strophic measurements reveal an energetic mesoscale ADCP bin (centered at 28 m), the velocity of the jet core 2 field, the in situ observations presented next reveal is 1.85 m s 1, with asymmetric lateral shear across the much larger values of velocity and its spatial gradients at front. The cyclonic shear on the cold eastern side of the submesoscale range. the jet, measured as dy/dx, has a maximum of 2.3f. Conversely, the anticyclonic shear flank of the jet peaks 5. Submesoscale processes at the front at 20.76f, as it is constrained from exceeding f (Shcherbina et al. 2013; Thomas et al. 2008). In contrast, a. Submesoscale frontal structure the satellite-derived vorticity reaches a peak magnitude In situ measurements augment the large-scale view of of 0.4f. The difference between satellite-measured geo- the eddy dipole from satellite ADT observations, re- strophic and in situ velocities reveals that while the vealing submesoscale characteristics embedded within satellite observations do a good job of representing the

Unauthenticated | Downloaded 10/06/21 12:07 PM UTC 1516 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 50

FIG. 3. Time–latitude Hovmöllers along a transect shown in Fig. 2 during August–September reveal the evolution of (a) SST; (b) along-transect gradient in SST, which highlights the thermal fronts between the eddies and the EAC; 2 and (c) FSLE (day 1), which characterizes the strain field over time. The black dashed line denotes the 12 Sep when the in situ measurements were taken. For consistency and time coverage, the SST used here is the objectively analyzed daily blended L4 SST 1-km resolution product (PO.DAAC g1sst product: https://podaac-opendap.jpl.nasa.gov/). mesoscale, finescale structure at the front is not captured chlorophyll a in the cyclonic eddy (Fig. 5f). The front by altimetry. tilts with depth in the direction of the anticyclonic Based on a high-resolution vertical slice from the eddy. Below the isopycnal outcrop the weakly strati- Triaxus, the front has a width of ;15 km, and extends fied, cold and fresh surface water of the cyclonic eddy down to 200-m depth (Fig. 5). The mixed layer base appears to subduct into the pycnocline (e.g., Fig. 5c). ascends from ;275 to ;175 m over a range of 60 km. Undulating isopycnals at the base of the mixed layer The front is demarcated by exceptionally strong hor- are the signatures of internal gravity waves with am- izontal gradients in all measured properties, with plitudes as large as 50 m and wavelengths ;15 km (if warmer, saltier, low-oxygen content water in the observations can be assumed near-instantaneous with deeper mixed layer of the anticyclonic eddy, and respect to the propagating internal waves), and their colder, fresher, high-oxygen content in the cyclonic impact is seen on all variables, particularly chlorophyll a eddy. There is a significantly higher concentration of and oxygen (Figs. 5c,f). Without additional information

Unauthenticated | Downloaded 10/06/21 12:07 PM UTC MAY 2020 A R C H E R E T A L . 1517

FIG. 4. (a) A VIIRS L2 swath measurement of SST at ;1542 UTC 12 Sep, the day the ship transected the front. The black arrows show ADCP-derived horizontal currents at 28 m. (b) The horizontal gradient of the SST field. The white arrows show shipboard-measured wind velocity, which is upfront at this time. The local Cartesian coordinates used to calculate derivates is shown. (c) Atmospheric forcing during in situ sampling; the total heat flux QHF (negative indicates the ocean loses heat) and Ekman equivalent heat flux QEQ (positive indicates upfront winds with a stabilizing influence). (d) In situ underway temperature measurement at 2 m and the SST interpolated onto the transect. (e) The along-transect gradient of the underway temperature and VIIRS SST. (f) Meridional y and zonal u velocity components. (g) Relative vorticity normalized by jfj, for ADCP-derived currents (red), and ADT-derived geostrophic currents interpolated onto the transect (blue). (SST data source: http://oceandata.sci.gsfc.nasa.gov). it is not clear whether these waves were generated re- Outside of the narrow front, the weaker shear is pre- motely or by the front itself (e.g., Shakespeare and dominantly ageostrophic with large gTWB. Such differ- Taylor 2014). ences are expected in the mixed layer and pycnocline The S-ADCP vertical profiles of the horizontal cur- where internal waves are evident, but not the ageo- rents reveal the outcropping isopycnals are associated strophic shear at the base of the front at 250-m depth, 2 with strong alongfront velocity, greater than 1 m s 1 all 40 km along track (Fig. 6f); in the same region the 2 the way down to the base of the mixed layer (Fig. 6a). across-front velocity is large (;25 cm s 1) and indicates This velocity has significant vertical shear (Figs. 6c,f), the water moving in the direction of the warm core eddy and is primarily in thermal wind balance along the front (Fig. 6b). The relative vorticity is asymmetric and until the base of the mixed layer, as shown in Fig. 7b, dominated by the cyclonic component along the front, where the departure from thermal wind balance is and coincident with high strain values, typical of sub- shown (in %) (Holmes et al. 2014): mesoscale motions (Figs. 6d,e)(Mahadevan 2016). The magnitude of horizontal buoyancy gradient M2 at y 2 2 dy d g ; 3 7 2 j – j the front is 6 10 s (Fig. 5e). This is very high, and g 5 dz dz 3 100. (13) comparable to strong fronts measured elsewhere [in the TWB dy jdyj 1 j gj ACC, Adams et al. (2017); in the Kuroshio, D’Asaro et al. dz dz (2011)]. From 4 months of glider measurements in the

Unauthenticated | Downloaded 10/06/21 12:07 PM UTC 1518 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 50

21 FIG. 5. Cross-front (x–z plane) sections measured by the Triaxus: (a) Conservative Temperature (8C), (b) Absolute Salinity (g kg ), 2 and (c) dissolved oxygen (mmol). Derived variables are (d) N2 the Brunt–Väisälä frequency (s 2), (e) M2 the horizontal density gradient 2 2 2 (s 2; note f 1M2 is thermal wind balance), and (f) chlorophyll a (mg m 3, as derived from fluorescence). Line contours represent potential 2 2 density referenced to the surface, from 1025.7 to 1026.7 kg m 3 in intervals of 0.035 kg m 3. In (a), (b), and (e), data from the underway thermosalinograph at 2-m depth has been plotted.

ACC, Viglione et al. (2018) found M2 magnitudes ex- derived strain rate increase at this time (Fig. 3c). 2 2 ceeded 6 3 10 7 s 2 less than 1% of the time (although Note in our calculation we use the full velocity rather this estimate is conservative given their 5-km sampling than the geostrophic component, which may include resolution and assuming the glider rarely sampled exactly strain from other sources, such as internal waves. perpendicular to fronts). When we calculate M2 at 2-m However, similar values can be obtained by taking the depth using thermosalinograph measurements with a ADT-derived geostrophic strain (;0.4f) and average 2 2 5-min sampling rate and a mean spatial resolution of 1 km, M2 at the front (;5 3 10 7 s 2), giving F 5 1.7 3 2 2 2 2 we obtain an even greater value of 1 3 10 6 s 2 (Fig. 5e). 10 17 s 5. This indicates horizontal strain is acting to The frontogenesis function has values of 2.5 3 increase the horizontal density gradient of the front, 2 2 10 17 s 5 (Fig. 7a), coincident with the altimetry- reaching a similar magnitude to that observed at a Gulf

Unauthenticated | Downloaded 10/06/21 12:07 PM UTC MAY 2020 A R C H E R E T A L . 1519

21 FIG. 6. Cross-front (x–z plane) sections measured by the shipboard ADCP: (a) alongfront velocity (m s ), (b) across-front velocity 2 2 2 (m s 1), (c) vertical shear of alongfront velocity (s 1), (d) vertical relative vorticity (s 1), (e) strain, and (f) vertical shear of across-front 2 2 velocity (s 1). Line contours represent potential density referenced to the surface, from 1025.7 to 1026.7 kg m 3 in intervals of 2 0.035 kg m 3.

Stream front in Thomas et al. (2013). Fluid parcels being Holmes et al. (2014) near the equator—convergence of carried along the front in the confluent region must ac- water at an upper-ocean front drives subduction of water, celerate to remain in cross-frontal geostrophic balance. which feeds divergent flow at depth that can lead to front- According to the semigeostrophic momentum balance olysis (although alongfront convergence/divergence can also such an acceleration is balanced by an ageostrophic absorb this subducting flow). secondary circulation (ASC; Hoskins 1982). This ageo- b. Subduction strophic flow will tend to subduct water below the front from the cold cyclonic side (e.g., see Fig. 12 of Pollard and Regier Subduction of water appears to be occurring at 200-m 1992). At the base of the front, there is a region of frontolysis depth, 55 km along track (Figs. 5–8). This is a region of in proximity to the subducting water parcel, associated with anticyclonic vorticity (Fig. 6d), which leads to near-zero slightly divergent flow. This process has been noted by absolute vorticity (defined here as za 5 f 1 ›y/›x) in this

Unauthenticated | Downloaded 10/06/21 12:07 PM UTC 1520 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 50

that ITEs can evolve from subducted water, which sep- arates from the mixed layer and moves into the ther- mocline, with anticyclonic vorticity, low-magnitude PV, and water properties of the mixed layer (i.e., high oxy- gen, low stratification, and T/S anomalies to the ambient thermocline water). The water parcel we observe has the same properties as the cyclonic eddy mixed layer, but with the opposite sign of vorticity; from this we can infer that the parcel gained anticyclonic vorticity as it sub- ducted (Fig. 6d). If PV is conserved, then the generation of anticyclonic vorticity can occur via vortex squashing (Spall 1995), or by differential vertical motions that drive vortex tilting of the horizontal vorticity associated with the alongfront vertical shear (Thomas 2008; Holmes et al. 2014). The subducting water parcel in our observations bears a striking resemblance to the model results of Thomas (2008, their Fig. 5). The generation of a coherent ITE is not guaranteed to occur for all subducting water parcels because mesoscale straining can disrupt the process by shearing apart the nascent subducting water parcel in elongated ribbons of low PV (see Thomas (2008), their Fig. 4). This may be the case here, but we cannot tell from our transient viewpoint. There are several mechanisms that can drive frontal subduction of surface water into the thermocline. In the idealized simulations of Spall (1995) and Thomas (2008), it is baroclinic instability of the frontal jet that drives vertical circulation. As the jet meanders and generates eddies, it generates confluent (divergent) flow associated with the downstream (upstream) side of the meander crests, that compresses (expands) the hori- zontal density gradient. The alongfront velocity accel- erates (decelerates) to remain in thermal wind balance, which is compensated for by an eddy overturning cell, leading to downwelling on the denser side and up- FIG. 7. Cross-front (x–z plane) sections of variables derived from welling on the lighter side. However, Spall (1997) both hydrographic and velocity measurements. (a) Frontogenesis function, superimposed in gray lines is the Triaxus profile track. showed that when the front is subjected to a large-scale

(b) Percent error of thermal wind balance gTWB; if the flow is in deformation field, the subduction rate is dominated by perfect thermal wind balance, gTWB will be zero. Cells not statis- the resulting frontal ASC rather than time-dependent tically significant have been blanked out. (c) Ertel potential vor- upwelling/downwelling associated with meandering. ticity q. In (a) and (b), line contours represent potential density 2 More recently, Brannigan et al. (2017) showed in an referenced to the surface, from 1025.7 to 1026.7 kg m 3 in intervals 2 of 0.035 kg m 3. In (c), line contours represent the mixed layer idealized model that subduction only existed in their 2 2 PV 5 5 3 10 10 s 3. The black box denotes the zoomed-in region highest-resolution model runs (0.25 and 2 km, and not shown in Fig. 8. 4 km), and was directly related to submesoscale instabilities. From the observations alone, we cannot state con- 2 2 water parcel as f ’28 3 10 5 s 1. This water parcel also clusively which mechanism is driving the subduction, or exhibits low stratification, low-magnitude PV (Fig. 7c), if different mechanisms are occurring simultaneously and is associated with high oxygen and chlorophyll a (Spall 1997). However, given the large-scale strain ob- content (Figs. 5c,f). The low-magnitude PV water parcel is served, we expect that strain-induced frontogenesis and separated from the mixed layer by a band of negative PV the resultant ASC plays a central role. Vertical velocity (i.e., white parcel in Fig. 7c). This is characteristic of the scales as W ; RodU, where Ro is the Rossby number formation mechanism of ITEs, as shown in the simulations (z/f), d is the ratio of depth to length scales (D/L), of Spall (1995) and Thomas (2008). Their simulations show and U is the characteristic velocity (Mahadevan 2016).

Unauthenticated | Downloaded 10/06/21 12:07 PM UTC MAY 2020 A R C H E R E T A L . 1521

FIG. 8. (a),(b) A 1D profile of PV in distance and depth, respectively. Blue lines denote the confidence intervals. (c) A zoom-in of Ertel PV q, the region denoted by the black box in Fig. 7c. Line contours represent the mixed layer 210 23 PV 5 5 3 10 s . (d),(e) 1D profiles of qbc (green line) and qvert (blue line). The gray shaded bands denote the region of positive PV, which is inherently unstable.

Based on the observations of the water parcel here, (fq , 0) is inherently unstable (Hoskins 1974). A 2 W ; 250 m day 1 (for Ro 5 1, D 5 50 m, L 5 15 km, U 5 zoomed-in view of the area (Fig. 8)revealsthatthis 2 1ms 1), signifying very strong vertical motions associ- positive PV occurs due to a stable but weak N2,such 2 ated with the front. Based on asymmetric vorticity dis- that qvert is weak, and a strong M , indicating the pre- tribution, we would expect the vertical velocity to dominance of the baroclinic PV component qbc in gen- exhibit stronger downwelling velocities over a smaller erating fq , 0[section 3c(2)]. We calculated PV in both area in the cyclonic zone, compared with broader, Cartesian and cylindrical coordinates, and found an av- slower upwelling in the anticyclonic zone (Mahadevan erage percent difference of 9% [we show PV calculated in and Tandon 2006). Previous studies calculated the cylindrical coordinates, per Eq. (9),inFigs. 7 and 8]. PV Omega equation (e.g., Rudnick 1996),butwehavenot with the opposite sign of f due to a strong M2 and weak N2 done so considering the degree of ageostrophy exhibited signifies the potential for symmetric instability, as op- (Fig. 7b), and that both quasigeostrophic (QG) and posed to gravitational or inertial instabilities (Thomas surface QG omega equations fail to capture Rossby et al. 2013). Symmetric instability, also termed slantwise number O(1) submesoscale dynamics (Mahadevan convection (Taylor and Ferrari 2009), produces a slanted 2016). In the following sections we show the presence overturning cell at the base of the front. of submesoscale instabilities that may have also played a Positive PV can only be created by surface forcing role in this subduction. across the isopycnal outcrop and is ephemeral since it initiates instabilities that act to mix it out and return the c. Submesoscale instability and atmospheric forcing flow to marginal stability. To investigate the source of There is a region of positive PV at the base of the positive PV observed at this front, we consider the at- front, in close proximity to the subducted parcel mospheric conditions around this time period (Fig. 4c (Fig. 7c). Positive PV in the Southern Hemisphere and the online supplemental material). The air–sea

Unauthenticated | Downloaded 10/06/21 12:07 PM UTC 1522 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 50

FIG. 9. High-resolution swath measurements of SST from VIIRS. These panels show the progression of the front before, during, and after in situ sampling. Frontal instabilities are observed in satellite imagery on 12–15 Sep, with a wavelength of ;30 km. The color map denotes the horizontal gradient in SST. The data are processed to level L2, with a spatial resolution of 750 m, obtained on 15 Sep 2017 from an ascending swath at approximately 0330 UTC from Suomi-NPP VIIRS. Data source: http://oceandata.sci.gsfc.nasa.gov. The bottom- right panel FSLE on 15 Sep shows good agreement to the high-resolution SST frontal images. Red crosses denote centers of the frontal instability highlighted by orange circles on days 14 and 15 Sep. heat flux, calculated from shipboard meteorological and positive PV. Our observations are taken just over two oceanographic measurements at the front, was destabi- inertial periods after downfront winds stopped (inertial 2 lizing, exhibiting ocean heat loss of up to 200 W m 2 to period at 358Sis;21 h). Taylor and Ferrari (2009) note the atmosphere across the front (Fig. 4c). Such heat the existence of a ‘‘fossil remnant mode’’ of symmetric fluxes can generate fq , 0(Haine and Marshall 1998), instability of longer wavelength that persists after the and influence frontogenesis and enhance subduction most-unstable mode has mostly driven PV back to zero. (Yoshikawa et al. 2001). However, when compared to We see most of the front is tilting toward restratification; the effect of wind forcing at the front, the heat flux is an it is only at the base where the front is still vertical and order of magnitude smaller. In Fig. 4c, the effect of wind characterized by strong M2 and weak N2. is represented as an Ekman equivalent heat flux (Q ). EK d. Baroclinic instability During in situ sampling of the front, winds were upfront, 22 with a positive QEK of ;2500 W m . Therefore, any Satellite imagery shows the presence of cyclonic water column destabilization via the surface heat flux eddies along the front 2 days later, on 14–15 September would have been counteracted by the much larger sta- (Fig. 9), with length scales of ;15 km (unstable wave- bilizing effect of wind forcing. length of ;30 km). The submesoscale horizontal length However, preceding the period of in situ sampling, the scale can be expressed in terms of the horizontal or wind was downfront (between 8 and 10 September; see vertical buoyancy gradients, as L 5 M2H/f2 or L 5 NH/f, supplemental material). Without recourse to additional respectively (Thomas et al. 2008). Using mean in situ observations, we can only speculate that this wind values of the buoyancy gradient at the front (M2 5 5 3 2 2 2 2 forcing may have generated the positive PV we ob- 10 7 s 2 and N2 5 5 3 10 5 s 2) and frontal depth H 5 served on the 12 September, which could be a remnant 200 m yields L 5 14 km (17 km), based on M2 (N2). of a larger region of positive PV that subsequently Thus, we expect these eddies begin as submesoscale mixed out. The PV we observed was at the base of the instabilities of the front, given that they evolved from a front, furthest from the influence of stabilizing wind. high Ro (and low Richardson) flow (Thomas et al. 2008), Indeed, the isopycnals at the base are more vertical, and with a length scale on the order of the mixed layer while higher in the water column they are more tilted as deformation scale. On 13 and 14 September, the wind would be expected from upfront winds. Mahadevan was strongly downfront and therefore would have had a (2016) note it can take several inertial periods to mix out destabilizing influence (supplemental material).

Unauthenticated | Downloaded 10/06/21 12:07 PM UTC MAY 2020 A R C H E R E T A L . 1523

FIG. 10. (a) Satellite-derived time–latitude Hovmöller plots, and (b) time standard deviation of the mesoscale SST gradient reveals strong eddy activity poleward of ;328S. (c) FSLE (spatially smoothed) on 4 Sep, superimposed by contour lines of the horizontal SST gradient shows the strong match of these values at frontal zones. The black line denotes the transect over which the Hovmöller and the correlation were taken; (d) Along-transect values of jFSLEj and j=SSTj as shown in (c); (e) along-transect correlation between jFSLEj and j=SSTj reveals event-type distribution of periods with high correlation, when anticyclonic eddy shedding events or eddy dipoles are active in the Tasman Sea.

Baroclinic instability can occur as frontal meandering formed as the anticyclonic EAC retroflection separated and eddy formation. Several modeling studies have from the main EAC branch, and a cyclonic eddy prop- shown how baroclinic instability of a front generates agated westward in between (section 4, Fig. 2). The cyclonic frontal eddies at the surface associated with eddies of the dipole had distinct water characteristics, subduction of the denser water (Spall 1995; Manucharyan and the geostrophic strain between their cores sharp- and Timmermans 2013;seesection 5b). However, when a ened the existing horizontal gradients (Fig. 3c). A strain large-scale deformation field is present, instead of the field generated by co-interacting mesoscale eddies is frontal jet devolving into a field of eddies, it remains known to generate submesoscale gradients of buoyancy quasistationary, oscillating between periods of meander (Klein and Lapeyre 2009). This situation preconditioned growth and decay as the mesoscale frontogenesis coun- the flow to frontogenesis, subduction and submesoscale teracts frontolysis by instabilities (Spall 1997). This is a instability (e.g., Haine and Marshall 1998). To investi- qualitatively similar situation to the frontal jet observed gate how typical the eddy dipole structure is, we sur- here that is quasistationary and coherent–at times straight veyed 25 years of satellite imagery in the region. We (undergoing mesoscale frontogenesis), and at another calculate the SST gradient along a transect that bisects time meandering (after destabilizing winds). Indeed, the dipole (›SST/›y, where y is the along-transect dis- Fig. 9 reveals the ongoing evolution of this front, which by tance; see Fig. 10). The SST gradient emphasizes the 18 September has shifted and strengthened (in tempera- location of fronts between warm and cold water masses ture gradient), presumably due to frontogenesis driven by and can be used as a proxy to detect horizontal density the mesoscale strain field of the dipole. gradients. Such mesoscale density gradients have been shown in a high-resolution model to be highly correlated with submesoscale vertical velocity (Rosso et al. 2015). 6. Discussion An eddy dipole will produce an alternating positive– negative pattern in ›SST/›y. Poleward propagating di- a. A persistent mesoscale eddy dipole pattern poles are evident in a time–latitude Hovmöller plot, The front was generated by strain between two characterized by downward slopes of alternating positive– counterrotating eddies—a mesoscale eddy dipole—that negative bands (Fig. 10a). There is also a clear distinction

Unauthenticated | Downloaded 10/06/21 12:07 PM UTC 1524 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 50

FIG. 11. The anomalous submesoscale lens observed by BR12. ADT maps superimposed with geostrophic currents on (a) 15 Jun 2010, (b) 16 Aug 2010, and (c) 5 Oct 2010, showing the evolution and persistence of the mesoscale anticyclone. (d) Chlorophyll a concentration 2 from the OC3 algorithm (mg m 3) and (e) glider measurements of oxygen along the track denoted by the black line in (c) and white line in (d). A subducting tongue of high oxygen water is evident at the periphery of the anticyclone—the glider sampling of this tongue feature is denoted by the red stop/start circles in (d). of flow characteristics upstream and downstream of the influence of mesoscale straining—which increases avail- mean EAC separation. Poleward of approximately 338S, able potential energy—and the stabilizing restratification there is significantly higher variance due to the larger eddy that occurs through rapid frontogenesis and atmospheric magnitudes, associated with eddy avenue (Everett et al. buoyancy fluxes that drive instabilities. Such competing 2012), and the characteristic positive–negative banding forces lead to periods of frontal meandering and sub- from eddy dipoles. The horizontal gradient of SST is cor- mesoscale instability, and periods when the front is related with the FSLE field (Figs. 10c–e), which we use to straighter (e.g., Fig. 9)(Spall 1997). diagnose the time-integrated effect of mesoscale strain, b. Submesoscale lenses of anomalous water in the and is associated with large buoyancy gradients (e.g., Tasman Sea Siegelman et al. 2020). For the dipole we sampled, the SST gradient and FSLE field correlate along the transect (r 5 Results presented here may explain the generation 0.75), as both metrics exhibit extrema at the frontal loca- mechanism of anomalous submesoscale lenses previously tions between the EAC and dipole eddies. Along-transect observed in the Tasman Sea (BR12)(Fig. 11). Brannigan correlation over the 25 years of satellite imagery exhibits et al. (2017) hypothesized that the lenses observed an event-type pattern, with short periods (days to weeks) by BR12 were formed by submesoscale instabilities at of high correlation between the two fields when strong the surface. We present direct measurements of sub- fronts are present in the region. mesoscale features along an ocean front that lead to the The persistence of fronts in this region invokes a pat- subduction of surface water into the thermocline of an tern of recurrent competition between the destabilizing anticyclonic eddy. BR12 highlight that all observed

Unauthenticated | Downloaded 10/06/21 12:07 PM UTC MAY 2020 A R C H E R E T A L . 1525 lenses were within anticyclonic eddies. This agrees with The strong vertical velocity associated with sub- the intrathermocline eddy generation mechanism; at a mesoscale eddy-driven subduction has important im- front between a warm and cold core eddy, the colder, plications for the export of biogeochemical tracers denser water subducts beneath the relatively lighter (heat, salt, carbon, dissolved gases, etc.) from the anticyclonic eddy. The lifespan of these dipoles is on the mixed layer below the pycnocline (Klein and Lapeyre order of weeks to months, so the potential for atmo- 2009; Omand et al. 2015; Mahadevan 2016). A recent spheric forced frontal instabilities and subduction is high. estimate of particulate organic carbon (POC) export Duringthesamecruiseweobservedananomalous showed that in regions of strong mesoscale eddy activity water lens at 600-m depth in the anticyclonic eddy, and seasonal MLD variability, up to 50% of POC export revealed by CTD measurements of oxygen, tempera- may occur through localized eddy-driven subduction ture, and salinity (supplemental material). The lens (Omand et al. 2015). In the Tasman Sea, chlorophyll a had higher oxygen content than ambient water, and concentration is anomalously high south of the EAC higher salinity and temperature. While it is difficult to (Everett et al. 2014), where mesoscale eddy dipoles en- discern its size based on seven scattered CTD casts, it ergize submesoscale frontal instability and subduction was not large enough to be sampled by all—only three (e.g., Fig. 7c). This coupling suggests submesoscale sampled the feature—so it probably has a horizontal eddy-driven subduction contributes to both primary scale less than 50 km, and vertical scale of ;200 m. This productivity and air–sea carbon dioxide exchange. closely matches the BR12 lenses of ;40 km and ;200 m. d. Open questions Bass Strait water (Fig. 1a) has been shown to flow northward into the Tasman Sea and move along the Our analysis is restricted to one high-spatial resolu- shelf for up to 200 km (BR12), during which time there tion section of a 3D time-dependent process, so many are many opportunities for it to interact with anticy- questions remain. There are several mechanisms that clonic eddies shed from the EAC. Shelf-deep ocean can drive or influence the subduction of surface water, interaction of this type has been documented in the including: baroclinic instability of a frontal jet and re- Mid-Atlantic Bight; Gulf Stream anticyclonic eddies sultant eddy overturning cell (e.g., Spall 1995), strain- come into contact with shelf water, which instigates driven frontogenesis and resultant secondary circulation frontogenesis and subduction (Zhang and Partida (e.g., Spall 1997), and submesoscale instabilities, such 2018). The submesoscale lenses observed by BR12 as symmetric instability (Brannigan 2016). To delin- had water properties matching Bass Strait. The cold eate the relative influence of these processes requires core eddy sampled in September 2017 originated from more high-spatial resolution and time-dependent ob- offshore in the Tasman Sea, so was unlikely to contain servations, which we do not have. Based on available Bass Strait water. Conversely, the BR12 mesoscale an- observations (Figs. 3 and 7a) and model evidence (Spall ticyclonic eddy that contained the lens was a persistent 1997), it seems likely that strain-induced frontogenesis feature close to the shelf throughout 2010, so had ample dominated, given the large strain values measured. opportunity to interact with shelf water (Fig. 11). However, different mechanisms can act simultaneously (McWilliams 2016), and with observational evidence for c. Implications mixed layer baroclinic instability (Fig. 9) and symmetric In situ measurements reveal the front was susceptible instability (Fig. 8) it is possible that each of these phys- to symmetric instability (SI). This has been demon- ical processes has an influence on the vertical exchange strated in other western boundary currents [e.g., in the of water at this front. More questions generated by our Gulf Stream, Thomas et al. (2013); in the Kuroshio, observations—but which require numerical models to D’Asaro et al. (2011)]. Unlike mesoscale baroclinic in- answer—include: What is the relationship between fq , 0 stability, which leads to an inverse cascade of KE, and ITE formation? What is the mechanism by which an SI extracts KE from the balanced ocean flow and ITE formed at an upper-ocean front moves down to provides a route for dissipation via small-scale turbu- depths of ;600 m? How does variability in wind speed lence. Such processes are an important link in the cas- and direction effect the stability of a front in the presence cade of energy from large scales where it is input to the of mesoscale strain and submesoscale instabilities? small-scale where it is dissipated (Molemaker et al. 2010; McWilliams 2016). Observations presented here pro- 7. Summary vide the first empirical evidence that the EAC system is likely to comprise this instability type, which could play In situ measurements at a strong upper-ocean front re- an important role in the energy balance of the Tasman veal the existence of submesoscale variability embedded Sea ocean circulation. within the mesoscale circulation, south of the EAC

Unauthenticated | Downloaded 10/06/21 12:07 PM UTC 1526 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 50 separation zone. This front appeared to be undergoing and distributed by Aviso1 (https://www.aviso.altimetry.fr/) strain-induced frontogenesis and was associated with with support from CNES. Ocean color and SST imagery subduction of surface water beneath the mixed layer. was produced by NASA’s Ocean Biology Processing The subducting water parcel had low PV, high oxygen, Group (https://oceandata.sci.gsfc.nasa.gov/). We make and anticyclonic vorticity, providing observational evi- use of ‘‘cmocean’’ colorbars developed by Kristen Thyng, dence for the formation of an intrathermocline eddy at and ‘‘brewermap’’ colorbars developed by Cynthia Brewer an upper-ocean front, as predicted by idealized numer- and colleagues. This research was partially supported by an ical simulations (Spall 1995; Thomas 2008). At the base Australian Research Council Linkage Project Grant to of the front, measurements suggest the flow was sus- MR (LP150100064). We are thankful to the two reviewers ceptible to symmetric instability (with fq , 0) due to a whose insightful feedback helped improve the quality this very large horizontal buoyancy gradient with a weak but manuscript. stable vertical buoyancy gradient. We postulate that mesoscale frontogenesis preconditioned the flow to SI, REFERENCES and subsequent downfront winds destabilized the flow, é injecting positive PV into the frontal outcrop. Two days Adams, K. A., P. Hosegood, J. R. Taylor, J. B. Sall e, S. Bachman, R. Torres, and M. Stamper, 2017: Frontal circulation and after the front was sampled, cloud-free satellite imagery submesoscale variability during the formation of a Southern reveals beautiful submesoscale eddies along the front, Ocean mesoscale eddy. J. Phys. Oceanogr., 47, 1737–1753, with a length scale of ;15 km (Fig. 9). https://doi.org/10.1175/JPO-D-16-0266.1. The frontal subduction of surface water observed here Archer, M. R., L. K. Shay, B. Jaimes, and J. Martinez-Pedraja, may explain the existence of anomalous submesoscale 2015: Observing frontal instabilities of the Florida Current ; using high frequency radar. Coastal Ocean Observing lenses observed by BR12 at 600-m depth in the cores Systems,Y.Liu,H.Kerkering,andR.H.Weisberg,Eds., of several anticyclonic eddies. These lenses appear to be Elsevier, 179–208, https://doi.org/10.1016/B978-0-12-802022- relatively common, since we identified a similar feature 7.00011-0. in the core of the anticyclonic eddy we sampled (sup- ——, M. Roughan, S. R. Keating, and A. Schaeffer, 2017: On the plemental material). Twenty-five years of satellite im- variability of the East Australian Current: Jet structure, meandering, and influence on shelf circulation. J. Geophys. Res. agery reveals the existence of strong mesoscale strain Oceans, 122, 8464–8481, https://doi.org/10.1002/2017JC013097. coupled with strong temperature fronts in this region ——, S. R. Keating, M. Roughan, W. E. Johns, R. Lumkpin, and indicates that the conditions which led to frontal F. Beron-Vera, and L. K. Shay, 2018: The kinematic similarity subduction observed here are a persistent feature. of two western boundary currents revealed by sustained high- Our findings suggest that the energetic mesoscale eddy resolution observations. Geophys. Res. Lett., 45, 6176–6185, https://doi.org/10.1029/2018GL078429. field south of the EAC separation zone is a hotspot for Baird, M. E., and K. R. Ridgway, 2012: The southward transport of submesoscale instability and subduction, and generate sub-mesoscale lenses of Bass Strait Water in the centre of anti- questions about the relationship between submesoscale cyclonic mesoscale eddies. Geophys. Res. Lett., 39, L02603, and mesoscale circulations, their space–time variability, https://doi.org/10.1029/2011GL050643. net impact on the export of surface water into the ocean Boccaletti, G., R. Ferrari, and B. Fox-Kemper, 2007: Mixed layer instabilities and restratification. J. Phys. Oceanogr., 37, 2228– interior, and influence on biological productivity. 2250, https://doi.org/10.1175/JPO3101.1. Brannigan, L., 2016: Intense submesoscale upwelling in anticy- Acknowledgments. We heartily thank the scientists, clonic eddies. Geophys. Res. Lett., 43, 3360–3369, https:// Australian Marine National Facility staff and crew of the doi.org/10.1002/2016GL067926. R/V Investigator who made the voyage a success, under the ——, D. P. Marshall, A. C. Naveira Garabato, A. G. Nurser, and J. Kaiser, 2017: Submesoscale instabilities in mesoscale eddies. guidance of chief scientist Iain Suthers, master John Highton, J. Phys. Oceanogr., 47, 3061–3085, https://doi.org/10.1175/ and voyage manager Don McKenzie. Special thanks JPO-D-16-0178.1. to Paulina Cetina-Heredia, Carlos Rocha, and Eduardo Bull, C. Y. S., A. E. Kiss, N. C. Jourdain, M. H. England, and E. van Queiroz for their vital role in planning the instrument Sebille, 2017: Wind forced variability in eddy formation, eddy deployment, and Matt Boyd, Karl Malakoff, Jason Fazey, shedding, and the separation of the East Australian Current. J. Geophys. Res. Oceans, 122, 9980–9998, https://doi.org/ Sven Gastauer, and Kendal Sherrin for technical support and 10.1002/2017JC013311. making the 2am–2pm shift so much fun. All data collected Callies, J., R. Ferrari, J. M. Klymak, and J. Gula, 2015: Seasonality onthecruiseareavailablethrough the Australian Ocean in submesoscale turbulence. Nat. Commun., 6, 6862, https:// Data Network (www.aodn.org.au). The SSALTO/DUACS doi.org/10.1038/ncomms7862. altimeter products were produced and distributed by Capet, X., J. C. McWilliams, M. J. Molemaker, and A. F. Shchepetkin, 2008: Mesoscale to submesoscale transition in the Copernicus Marine and Environment Monitoring the California Current System. Part I: Flow structure, eddy Service (http://www.marine.copernicus.eu), the FSLE flux, and observational tests. J. Phys. Oceanogr., 38, 29–43, were provided in collaboration with LOcean and CTOH https://doi.org/10.1175/2007JPO3671.1.

Unauthenticated | Downloaded 10/06/21 12:07 PM UTC MAY 2020 A R C H E R E T A L . 1527

Cetina-Heredia, P., M. Roughan, E. van Sebille, and M. A. Holmes,R.M.,L.N.Thomas,L.Thompson,andD.Darr,2014: Coleman, 2014: Long-term trends in the East Australian Potential vorticity dynamics of tropical instability vortices. J. Phys. Current separation latitude and eddy driven transport. Oceanogr., 44, 995–1011, https://doi.org/10.1175/JPO-D-13-0157.1. J. Geophys. Res. Oceans, 119, 4351–4366, https://doi.org/ Hoskins, B. J., 1974: The role of potential vorticity in symmetric 10.1002/2014JC010071. stability and instability. Quart. J. Roy. Meteor. Soc., 100, 480– Chereskin, T. K., C. B. Rocha, S. T. Gille, D. Menemenlis, and 482, https://doi.org/10.1002/qj.49710042520. M. Passaro, 2019: Characterizing the transition from balanced ——, 1982: The mathematical theory of frontogenesis. Annu. to unbalanced motions in the southern California Current. Rev. Fluid Mech., 14, 131–151, https://doi.org/10.1146/ J. Geophys. Res. Oceans, 124, 2088–2109, https://doi.org/ annurev.fl.14.010182.001023. 10.1029/2018JC014583. Jungclaus, J. H., 1999: A three-dimensional simulation of the for- D’Asaro, E. A., 1988: Generation of submesoscale vortices: A new mation of anticyclonic lenses (meddies) by the instability of an mechanism. J. Geophys. Res., 93, 6685–6693, https://doi.org/ intermediate depth boundary current. J. Phys. Oceanogr., 29, 10.1029/JC093iC06p06685. 1579–1598, https://doi.org/10.1175/1520-0485(1999)029,1579: ——, C. Lee, L. Rainville, R. Harcourt, and L. Thomas, 2011: ATDSOT.2.0.CO;2. Enhanced turbulence and energy dissipation at ocean fronts. Kim, S. Y., and Coauthors, 2011: Mapping the US west coast sur- Science, 332, 318–322, https://doi.org/10.1126/science.1201515. face circulation: A multiyear analysis of high-frequency radar d’Ovidio, F., V. Fernández, E. Hernández-García, and C. López, observations. J. Geophys. Res., 116, C03011, https://doi.org/ 2004: Mixing structures in the Mediterranean Sea from finite- 10.1029/2010JC006669. size Lyapunov exponents. Geophys. Res. Lett., 31, L17203, Klein, P., and G. Lapeyre, 2009: The oceanic vertical pump induced by https://doi.org/10.1029/2004GL020328. mesoscale and submesoscale turbulence. Annu. Rev. Mar. Sci., 1, Dugan, J. P., R. P. Mied, P. C. Mignerey, and A. F. Schuetz, 1982: 351–375, https://doi.org/10.1146/annurev.marine.010908.163704. Compact, intrathermocline eddies in the Sargasso sea. J. Geophys. Klymak, J. M., and Coauthors, 2016: Submesoscale streamers Res., 87, 385–393, https://doi.org/10.1029/JC087iC01p00385. exchange water on the north wall of the Gulf Stream. Everett, J. D., M. E. Baird, P. R. Oke, and I. M. Suthers, 2012: An Geophys. Res. Lett., 43, 1226–1233, https://doi.org/10.1002/ avenue of eddies: Quantifying the biophysical properties of 2015GL067152. mesoscale eddies in the Tasman Sea. Geophys. Res. Lett., 39, Kostianoy, A., and I. Belkin, 1989: A survey of observations on in- L16608, https://doi.org/10.1029/2012GL053091. trathermocline eddies in the world ocean. Mesoscale/Synoptic ——, ——, M. Roughan, I. M. Suthers, and A. M. Doblin, 2014: Coherent structures in Geophysical Turbulence,J.Nihouland Relative impact of seasonal and oceanographic drivers on B. Jamart, Eds., Elsevier Oceanography Series, Vol. 50, Elsevier, surface chlorophyll a along a Western Boundary Current. 821–841, https://doi.org/10.1016/S0422-9894(08)70223-X. Prog. Oceanogr., 120, 340–351, https://doi.org/10.1016/ Lee, C. M., L. N. Thomas, and Y. Yoshikawa, 2006: Intermediate water j.pocean.2013.10.016. formation at the Japan/East Sea subpolar front. Oceanography, ——, H. S. Macdonald, M. E. Baird, J. Humphries, M. Roughan, 19, 110–121, https://doi.org/10.5670/oceanog.2006.48. and I. M. Suthers, 2015: Cyclonic entrainment of precondi- Macdonald, H. S., M. Roughan, M. E. Baird, and J. Wilkin, 2013: A tioned shelf waters into a Frontal Eddy. J. Geophys. Res. numerical modeling study of the East Australian Current en- Oceans, 120, 677–691, https://doi.org/10.1002/2014JC010301. circling and overwashing a warm-core eddy. J. Geophys. Res. Fairall, C. W., E. F. Bradley, D. P. Rogers, J. B. Edson, and G. S. Oceans, 118, 301–315, https://doi.org/10.1029/2012JC008386. Young, 1996: Bulk parameterization of air-sea fluxes for Mahadevan, A., 2016: The impact of submesoscale physics on tropical ocean-global atmosphere coupled-ocean atmo- primary productivity of plankton. Annu. Rev. Mar. Sci., 8, 161– sphere response experiment. J. Geophys. Res., 101, 3747–3764, 184, https://doi.org/10.1146/annurev-marine-010814-015912. https://doi.org/10.1029/95JC03205. ——, and A. Tandon, 2006: An analysis of mechanisms for sub- Fischer, J., P. Brandt, M. Dengler, M. Müller, and D. Symonds, mesoscale vertical motion at ocean fronts. Ocean Modell., 14, 2003: Surveying the upper ocean with the Ocean Surveyor: A 241–256, https://doi.org/10.1016/j.ocemod.2006.05.006. new phased array Doppler current profiler. J. Atmos. Oceanic Malan, N., and Coauthors, 2020: Eddy-driven cross-shelf transport Technol., 20, 742–751, https://doi.org/10.1175/1520-0426(2003) in the East Australian Current separation zone. J. Geophys. 20,742:STUOWT.2.0.CO;2. Res. Oceans, 125, e2019JC015613, https://doi.org/10.1029/ Flament, P., L. Armi, and L. Washburn, 1985: The evolving 2019JC015613. structure of an upwelling filament. J. Geophys. Res., 90, Mantovanelli, A., S. Keating, L. R. Wyatt, M. Roughan, and 11 765–11 778, https://doi.org/10.1029/JC090iC06p11765. A. Schaeffer, 2017: Lagrangian and Eulerian characterization Fox-Kemper, B., R. Ferrari, and R. Hallberg, 2008: Parameterization of two counter-rotating submesoscale eddies in a western of mixed layer eddies. Part I: Theory and diagnosis. J. Phys. boundary current. J. Geophys. Res. Oceans, 122, 4902–4921, Oceanogr., 38, 1145–1165, https://doi.org/10.1175/2007JPO3792.1. https://doi.org/10.1002/2016JC011968. Godfrey, J. S., G. R. Cresswell, T. J. Golding, A. F. Pearce, and Manucharyan, G. E., and M. L. Timmermans, 2013: Generation R. Boyd, 1980: The separation of the east Australian current. and separation of mesoscale eddies from surface ocean fronts. J. Phys. Oceanogr., 10, 430–440, https://doi.org/10.1175/1520- J. Phys. Oceanogr., 43, 2545–2562, https://doi.org/10.1175/ 0485(1980)010,0430:TSOTEA.2.0.CO;2. JPO-D-13-094.1. Gula, J., T. M. Blacic, and R. E. Todd, 2019: Submesoscale co- McDougall, T. J., and P. M. Barker, 2011: Getting started with herent vortices in the Gulf Stream. Geophys. Res. Lett., 46, TEOS-10 and the Gibbs Seawater (GSW) Oceanographic 2704–2714, https://doi.org/10.1029/2019GL081919. Toolbox. SCOR/IAPSO WG127, 28 pp., http://www.teos- Haine, T. W., and J. Marshall, 1998: Gravitational, symmetric, and 10.org/pubs/Getting_Started.pdf. baroclinic instability of the ocean mixed layer. J. Phys. Oceanogr., McWilliams, J. C., 1985: Submesoscale, coherent vortices in the 28, 634–658, https://doi.org/10.1175/1520-0485(1998)028,0634: ocean. Rev. Geophys., 23, 165–182, https://doi.org/10.1029/ GSABIO.2.0.CO;2. RG023i002p00165.

Unauthenticated | Downloaded 10/06/21 12:07 PM UTC 1528 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 50

——, 2016: Submesoscale currents in the ocean. Proc. Roy. Soc., Shakespeare, C. J., 2016: Curved density fronts: Cyclogeostrophic 472A, 20160117, https://doi.org/10.1098/rspa.2016.0117. adjustment and frontogenesis. J. Phys. Oceanogr., 46, 3193– Molemaker, M., J. McWilliams, and X. Capet, 2010: Balanced 3207, https://doi.org/10.1175/JPO-D-16-0137.1. and unbalanced routes to dissipation in an equilibrated Eady ——, and J. R. Taylor, 2014: The spontaneous generation of inertia– flow. J. Fluid Mech., 654, 35–63, https://doi.org/10.1017/ gravity waves during frontogenesis forced by large strain: Theory. S0022112009993272. J. Fluid Mech., 757, 817–853, https://doi.org/10.1017/jfm.2014.514. ——, ——, and W. K. Dewar, 2015: Submesoscale instability and Shcherbina, A. Y., E. A. D’Asaro, C. M. Lee, J. M. Klymak, M. J. generation of mesoscale anticyclones near a separation of the Molemaker, and J. C. McWilliams, 2013: Statistics of vertical California Undercurrent. J. Phys. Oceanogr., 45, 613–629, vorticity, divergence, and strain in a developed submesoscale https://doi.org/10.1175/JPO-D-13-0225.1. turbulence field. Geophys. Res. Lett., 40, 4706–4711, https:// Munk, W., L. Armi, K. Fischer, and F. Zachariasen, 2000: Spirals doi.org/10.1002/grl.50919. on the sea. Proc. Roy. Soc. London, 456A, 1217–1280, https:// ——, and Coauthors, 2015: The LatMix summer campaign: doi.org/10.1098/rspa.2000.0560. Submesoscale stirring in the upper ocean. Bull. Amer. Meteor. Oke, P. R., and Coauthors, 2019: Revisiting the circulation of Soc., 96, 1257–1279, https://doi.org/10.1175/BAMS-D-14-00015.1. the East Australian Current: Its path, separation, and eddy Siegelman, L., P. Klein, P. Riviere, A. F. Thompson, H. S. Torres, field. Prog. Oceanogr., 176, 102139, https://doi.org/10.1016/ M. Flexas, and D. Menemenlis, 2020: Enhanced upward heat j.pocean.2019.102139. transport at deep submesoscale ocean fronts. Nat. Geosci., 13, Omand, M. M., E. A. D’Asaro, C. M. Lee, M. J. Perry, N. Briggs, 50–55, https://doi.org/10.1038/s41561-019-0489-1. I. Cetinic´, and A. Mahadevan, 2015: Eddy-driven subduction Spall, M. A., 1995: Frontogenesis, subduction, and cross-front ex- exports particulate organic carbon from the spring bloom. change at upper ocean fronts. J. Geophys. Res., 100, 2543– Science, 348, 222–225, https://doi.org/10.1126/science.1260062. 2557, https://doi.org/10.1029/94JC02860. Poje, A. C., and Coauthors, 2014: Submesoscale dispersion in ——, 1997: Baroclinic jets in confluent flow. J. Phys. Oceanogr., 27, the vicinity of the Deepwater Horizon spill. Proc. Natl. Acad. Sci. 1054–1071, https://doi.org/10.1175/1520-0485(1997)027,1054: USA, 111, 12 693–12 698, https://doi.org/10.1073/pnas.1402452111. BJICF.2.0.CO;2. Pollard, R. T., and L. A. Regier, 1992: Vorticity and vertical cir- Stone, P. H., 1970: On non-geostrophic baroclinic stability: Part culation at an ocean front. J. Phys. Oceanogr., 22, 609–625, II. J. Atmos. Sci., 27, 721–726, https://doi.org/10.1175/1520- https://doi.org/10.1175/1520-0485(1992)022,0609:VAVCAA. 0469(1970)027,0721:ONGBSP.2.0.CO;2. 2.0.CO;2. Su, Z., J. Wang, P. Klein, A. F. Thompson, and D. Menemenlis, Pujol, M. I., Y. Faugère, G. Taburet, S. Dupuy, C. Pelloquin, 2018: Ocean submesoscales as a key component of the global M. Ablain, and N. Picot, 2016: DUACS DT2014: The new heat budget. Nat. Commun., 9, 775, https://doi.org/10.1038/ multi-mission altimeter data set reprocessed over 20 years. s41467-018-02983-w. Ocean Sci., 12, 1067–1090, https://doi.org/10.5194/os-12-1067- Suthers, I. M., and Coauthors, 2011: The strengthening East 2016. Australian Current, its eddies and biological effects—An in- Qiu, B., T. Nakano, S. Chen, and P. Klein, 2017: Submesoscale troduction and overview. Deep-Sea Res. II, 58, 538–546, transition from geostrophic flows to internal waves in the https://doi.org/10.1016/j.dsr2.2010.09.029. northwestern Pacific upper ocean. Nat. Commun., 8, 14055, Taylor, J. R., and R. Ferrari, 2009: On the equilibration of a sym- https://doi.org/10.1038/ncomms14055. metrically unstable front via a secondary shear instability. J. Fluid Rocha, C. B., S. T. Gille, T. K. Chereskin, and D. Menemenlis, Mech., 622, 103–113, https://doi.org/10.1017/S0022112008005272. 2016: Seasonality of submesoscale dynamics in the Kuroshio Thomas, L. N., 2008: Formation of intrathermocline eddies at extension. Geophys. Res. Lett., 43, 11 304–11 311, https:// ocean fronts by wind-driven destruction of potential vorticity. doi.org/10.1002/2016GL071349. Dyn. Atmos. Oceans, 45, 252–273, https://doi.org/10.1016/ Rosso, I., A. M. Hogg, A. E. Kiss, and B. Gayen, 2015: Topographic j.dynatmoce.2008.02.002. influence on submesoscale dynamics in the Southern Ocean. ——, and C. M. Lee, 2005: Intensification of ocean fronts by down- Geophys. Res. Lett., 42, 1139–1147, https://doi.org/10.1002/ front winds. J. Phys. Oceanogr., 35, 1086–1102, https://doi.org/ 2014GL062720. 10.1175/JPO2737.1. Roughan, M., S. R. Keating, A. Schaeffer, P. Cetina Heredia, ——, A. Tandon, and A. Mahadevan, 2008: Submesoscale pro- C. Rocha, D. Griffin, R. Robertson, and I. M. Suthers, 2017: A cesses and dynamics. Ocean Modeling in an Eddying Regime, tale of two eddies: The biophysical characteristics of two Geophys. Monogr., Vol. 177, Amer. Geophys. Union, 17–38, contrasting cyclonic eddies in the East Australian Current https://doi.org/10.1029/177GM04. System. J. Geophys. Res. Oceans, 122, 2494–2518, https:// ——, J. R. Taylor, R. Ferrari, and T. M. Joyce, 2013: Symmetric doi.org/10.1002/2016JC012241. instability in the Gulf Stream. Deep-Sea Res. II, 91, 96–110, Rudnick, D. L., 1996: Intensive surveys of the Azores front: 2. https://doi.org/10.1016/j.dsr2.2013.02.025. Inferring the geostrophic and vertical velocity fields. Thompson, A. F., A. Lazar, C. Buckingham, A. C. Naveira J. Geophys. Res., 101, 16 291–16 303, https://doi.org/10.1029/ Garabato, G. M. Damerell, and K. J. Heywood, 2016: Open- 96JC01144. ocean submesoscale motions: A full seasonal cycle of mixed Sasaki, H., P. Klein, B. Qiu, and Y. Sasai, 2014: Impact of oceanic- layer instabilities from gliders. J. Phys. Oceanogr., 46, 1285– scale interactions on the seasonal modulation of ocean dy- 1307, https://doi.org/10.1175/JPO-D-15-0170.1. namics by the atmosphere. Nat. Commun., 5, 5636, https:// Torres,H.S.,P.Klein,D.Menemenlis,B.Qiu,Z.Su,J.Wang, doi.org/10.1038/ncomms6636. S. Chen, and L.-L. Fu, 2018: Partitioning ocean mo- Schaeffer, A., A. Gramoulle, M. Roughan, and A. Mantovanelli, tions into balanced motions and internal gravity waves: A 2017: Characterizing frontal eddies along the East Australian modeling study in anticipation of future space missions. Current from HF radar observations. J. Geophys. Res. Oceans, J. Geophys. Res. Oceans, 123, 8084–8105, https://doi.org/ 122, 3964–3980, https://doi.org/10.1002/2016JC012171. 10.1029/2018JC014438.

Unauthenticated | Downloaded 10/06/21 12:07 PM UTC MAY 2020 A R C H E R E T A L . 1529

Viglione, G. A., A. F. Thompson, M. M. Flexas, J. Sprintall, and S. Swart, Yoo, J. G., S. Y. Kim, and H. S. Kim, 2018: Spectral descriptions 2018: Abrupt transitions in submesoscale structure in southern of submesoscale surface circulation in a coastal region. Drake Passage: Glider observations and model results. J. Phys. J. Geophys. Res. Oceans, 123, 4224–4249, https://doi.org/ Oceanogr., 48, 2011–2027, https://doi.org/10.1175/JPO-D-17-0192.1. 10.1029/2016JC012517. Wang, D. P., C. N. Flagg, K. Donohue, and H. T. Rossby, 2010: Yoshikawa, Y., K. Akitomo, and T. Awaji, 2001: Formation Wavenumber spectrum in the Gulf Stream from shipboard process of intermediate water in baroclinic current under ADCP observations and comparison with altimetry mea- cooling. J. Geophys. Res., 106, 1033–1051, https://doi.org/ surements. J. Phys. Oceanogr., 40, 840–844, https://doi.org/ 10.1029/2000JC000226. 10.1175/2009JPO4330.1. Zhang, W. G., and J. Partida, 2018: Frontal subduction of the Mid- Waugh, D. W., and E. R. Abraham, 2008: Stirring in the global Atlantic Bight shelf water at the onshore edge of a warm-core surface ocean. Geophys. Res. Lett., 35, L20605, https://doi.org/ ring. J. Geophys. Res. Oceans, 123, 7795–7818, https://doi.org/ 10.1029/2008GL035526. 10.1029/2018JC013794.

Unauthenticated | Downloaded 10/06/21 12:07 PM UTC